sed script from vi commands

[jad]

i have a few vi commands that i'd like converting to something that can run on a command line, something like sed'd be good ...

i've got an html doc that needs cleaning up ... someone elses i must say, i hand code mine, but he's got about 50 docs, and i was just thinking if there was an easier way than 'vi'ing each doc and typing in the following commands.

vi <doc>.htm (yuck, not even html)
:g/></s//>^M</g (yes that is ctrl-v, ctrl-m)
:g/<[^>]*$/j (repeated lots of times ... j joins lines and puts a space before each join, although i don't need the space)
:g/<\/P[^>]*>/s///g
:g/ *>/s//>/g

hope that gives you the idea ...

Thanks

 

[Einstein47]

You know that you can just put all those commands in a buffer for vi to perform for you. This isn't sed but, I think it will work:

for fn in *.htm;do
vi $fn <<-EOF
:g/></s//>^M</g
:g/<[^>]*$/j
:g/<\/P[^>]*>/s///g
:g/ *>/s//>/g
:wq
EOF
done

I too would be interested in the sed method, but while we wait - please try this kludge. Einstein47
("If vegetarians eat vegetables, what do humanitarians eat?&quot

 

[jamisar]

try

:%!sed -e 's/aaa/bbb/g' vox clamantis in deserto.

 

[vgersh99]

for CLIediting in ed ex
do
man ${CLIediting}
done; vlad
+---------------------------+
|#include<disclaimer.h> |
+---------------------------+

 

[jamisar]

good vlad.
prepend edittool(don't remember the name)
and append vi to the list.
the ';' after 'done' will sure be mis-undestanded
vox clamantis in deserto.

 

[jad]

yeah, cheers for that.

1) can i do an equivalent of join in sed ... and if so, can it do them repeatedly for a given criteria
2) i can do substitutions in sed easy enough thankyou ...

i know i could write this thing in awk if i needed to, but i thought it'd be interesting if one of you guys could show me a sed equivalent.

 

[jad]

nawk '
{outstr=$0}
outstr ~ /<[^>]*$/ {
    while (match(outstr,&quot;<[^>]*$&quot;) > 0) {
        getline;
        outstr=sprintf(&quot;%s%s&quot;,outstr,$0);
    }
}
{
    gsub(&quot;>[ ]*<&quot;,sprintf(&quot;>\n<&quot;),outstr);
    gsub(&quot;<\/P[^>]*>&quot;,&quot;&quot;,outstr);
    gsub(&quot; *>&quot;,&quot;>&quot;,outstr);
    print outstr;
}' <file>.html

which works fine until you get a very long concatenation of strings in the while loop ...
i guess i could do with a 2 commands, one to split the >< first then one to join after.

e.g.

nawk '
{
    gsub(&quot;>[ ]*<&quot;,sprintf(&quot;>\n<&quot;)); 
    print;
}' <file>.html |
nawk '
{
    outstr=$0
}
outstr ~ /<[^>]*$/ {
    while (match(outstr,&quot;<[^>]*$&quot;) > 0) {
        getline;
        outstr=sprintf(&quot;%s%s&quot;,outstr,$0);
    }
}
{
    gsub(&quot;<\/P[^>]*>&quot;,&quot;&quot;,outstr);
    gsub(&quot; *>&quot;,&quot;>&quot;,outstr);
    print outstr;
}'[code]

which does work ...

however, back to my original question, can anyone do this nicely in sed? :)

 

[jamisar]

to join lines in vi:
:adr,adrj
in sed:
{
N
s/\n/ /
} vox clamantis in deserto.

 

[jad]

hmm ... got a sed version to work ... sort of, has same limitation with combining lines though ... they do get a bit long (and i need to do it as 2 passes) ... but i can't get it to be clean and put in \n's easily ...

sed 's/> *</>\n</' file.html

doesn't work for example.

btw. a cleaned up version of the nawk is available:

nawk '
{
    gsub(&quot;> *<&quot;,&quot;>\n<&quot;);
    print;
}' <file>.html |
nawk '
{
    outstr=$0
}
outstr ~ /<[^>]*$/ {
    while (match(outstr,&quot;<[^>]*$&quot;) > 0) {
        getline;
        outstr=sprintf(&quot;%s %s&quot;,outstr,$0);
    }
}
{
    gsub(&quot;<\/P[^>]*>&quot;,&quot;&quot;,outstr);
    gsub(&quot; *>&quot;,&quot;>&quot;,outstr);
    print outstr;
}'

the sed code i wrote went like this:

#!/bin/sh
sed '
:Begin
s/> *</>¬</
/<[^>]*$/ {
N
s/\n/ /
b Begin
}
s/<\/P[^>]*>//
s/ *>/>/
' ch09_01.htm | tr '¬' '\n'

i chose the ¬ character because it shouldn't really ever be in an html file

 

[bigoldbulldog]

Instead of piping to

tr '¬' '\n'

you could add into your sed script two consecutive lines as:

s/¬//g

The backslash allows the substitution to use the newline.

s/¬/\n/g

doesn't work unless you can snatch the newline from the pattern space somewhere (a type of substitution that wouldn't apply here anyway).
Cheers,
ND

bigoldbulldog@hotmail.com